Physic nut (Jatropha curcas L.), a tropical shrub with medicinal properties has not undergone efficient selection for high seed yield or oil production over the years. Predictions on seed yield of this oilseed plant were based on isolated observations that did not consider the reductions in plant growth due to higher planting densities. The objective of this study was to quantify the genetic progress of seed yield using mass selection in a physic nut population at production age, evaluated in the Amazonic tropical climate. Genetic variability for seed yield exists among J. curcas genotypes, denoting the possibility of genetic progress through plant selection. There was also a tendency for the genotypes to maintain their genetic superiority over the time. The estimated genetic progress from the selection of 10 superior genotypes was 0.67 kg.plant -1 , which, compared to the overall mean, corresponds to a selection gain of over 99%. The new mean of these plants was 1.35 kg.tree -1 , equivalent to seed yield of 2.25 tons.ha -1 . For breeding, a minimum number of 30 genotypes were considered to ensure the genetic progress over the generations. The estimated mean of a recombination unit composed of the 30 superior genotypes was 1.17 kg.plant -1 , equivalent to a seed yield 1.95 tons ha -1. Although accurate, the genetic progress estimated in this study does not appear to provide a qualitative increase in the yield of this oilseed. New breeding strategies may consider the use of crossings of divergent plants with better agronomic traits.